Skip to content
Back to news
Labs· 2 min read

Overseas cloud and GPU servers upgraded: 1 Gbps included, up to 20 Gbps

Outside Beijing and Ningxia, cloud and GPU servers now include 1 Gbps of shared peak bandwidth with a monthly allowance; dedicated bandwidth becomes an add-on.

From today, cloud and GPU servers outside Beijing and Ningxia get a network upgrade: every server includes 1 Gbps of shared peak bandwidth with a monthly traffic allowance, and you can go up to 20 Gbps when you need more.

Why we changed it

Our overseas servers used to follow Mbps-based bandwidth plans: you got exactly what you paid for, all the time. That suits some workloads, but often put customers in a bind — modest traffic most days, then a release, a live event or a big sync that needs many times the speed for a few hours. To cover those hours, they had to pay for a very wide pipe all year.

Most international cloud providers moved to high shared peaks plus traffic long ago. We looked at their traffic pricing, AWS included, together with our own data-centre and backbone costs, and decided to switch all overseas regions over.

How to choose now

  • Shared bandwidth (peak): 1 Gbps with a monthly allowance by default, upgradable to 2, 5, 10 or 20 Gbps. Right for websites, APIs, game servers, data sync and most other workloads.
  • Dedicated bandwidth: the old Mbps plans are now called dedicated bandwidth and remain as an add-on, for workloads that need steady, predictable speed, such as video-conference relays or private back-to-origin links.
  • Extra traffic: anything beyond the monthly allowance is billed by usage, with live prices on the product page.

Beijing and Ningxia work differently and keep dedicated-bandwidth pricing; cloud servers there still need dedicated bandwidth, while GPU servers may go without.

How much faster in practice

Take an online shop hosted in Hong Kong: on 10 Mbps dedicated, images slowed noticeably during promotions; on a 1 Gbps shared peak, the same traffic spike downloads large assets in a small fraction of the time, with no increase in the monthly network bill. For AI teams, pulling tens of gigabytes of model weights goes from a coffee break to a few minutes.

What about existing servers

New overseas servers get the new bandwidth automatically. For servers already running, choose a new shared bandwidth or adjust dedicated bandwidth under “Change bandwidth at renewal” on the service page: the period already paid for is unaffected, the next renewal is priced on the new plan, and the change applies once it is paid.

Not sure which combination costs least? Open a ticket or ask in live chat and we will work it out from your past traffic.

More news

  1. Group

    Singfung Group launches its new brand and website: one system for work that used to be scattered

    The new site is more than a front page. Labs and Capital each run their own back office, and every piece of business and every to-do — for customers, founders and our own teams — now happens on one platform.

  2. Labs

    Daily billing for GPU cloud servers

    NVIDIA RTX 4090, RTX 5090 and L40S can now be rented by the day — for short training runs, model trials and bursts of demand.

  3. Capital

    Singfung Capital makes its sixteenth investment

    Capital runs as an evergreen vehicle, backing early-stage companies in core software, applied AI and industrial digitalisation.